Epidemiology and Biostatistics Unit
What is the EBU?
The Epidemiology and Biostatistics Unit (EBU) is a centralized facility for epidemiological, biostatistical and biomathematical expertise to enhance biomedical and health research. The EBU was created in 1996 by the division of Preventive and Behavioral Medicine. Its original role was to provide epidemiological and statistical support for the division. Recently, the unit has expanded and is available to provide support and expertise to faculty in other divisions in the Department of Medicine.
The aims of the EBU are to enhance epidemiological and biostatistical research, and to establish and foster long-term collaborations among clinical and basic researchers, and biostatisticians and epidemiologists in the Department of Medicine. This collaboration will lead to an improvement in the aspects of study design, data management and quality control, and statistical analysis and modeling of research undertaken in the department. The unit can ensure that investigators have access to expert statistical and epidemiological consultation and support, and that the most appropriate and up-to-date methods are used.
The EBU members have strong training in methodology with extensive experience and interest in applications and collaboration. This allows us to carry out our own research program and interests, and provide expertise to the medical community.
We provide necessary tools and expertise in a number of areas:
- Efficient cost effective study designs
- Design of clinical, observational (including health survey analysis), and basic science (laboratory) studies.
- Appropriate power calculations (sample size estimates) that balance clinical and laboratory limitations, cost and resource limitations, and statistical power and efficiency. In cases where the packaged power programs cannot be manipulated for appropriate sample size determination, we have the ability to provide simulations of the analyses that will be done.
- As needed, we can provide database development (MS Access and Lotus Notes), programming for ScanTools software for optically scannable forms, data management, data entry, and quality control and assurance. We have flexible data transfer software that can link databases to multiple statistical packages (and vice versa) [DBMSCOPY, StatTransfer]
- Statistical analysis and modeling:
- We collaborate with the researcher to understand and present an accurate representation of the statistical and clinical evidence and its implications for medical practice and policy. We realize that many analyses of studies can be done with traditional statistical methodology applied to the data and do not go out of our way to complicate the analysis and presentation of data. But we also take into consideration the unique nature of each study and can determine what tools are necessary to avoid bias and to statistically represent the medical model. The statistical tools available [SAS version 8, Stata version 6, SPSS 9.0, S-Plus2000] and the expertise in the use of these tools allow a wide range of analytic methods and methodological and software development.
- We cover a wide range of analyses that can be applied to studies:
- Summary Statistics
- Exploratory graphical analyses; visual representation of quantitative data.
- Exploratory (hypothesis generating) analyses
- Statistical Hypothesis testing (Decision making)
- Statistical modeling: linear and nonlinear modeling; independent and correlated data; hierarchical (or nested/mixed) models.
- Appropriate weighting for health survey analysis
- Development of statistical tools if appropriate tools do not exist for a particular study/data set.
We emphasize the clinical interpretation of the data. Our goal is to understand the medical aspect of the scientific question so that design and analysis is NOT provided in a vacuum or by cookbook.
More EBU information, on financial support and contacts can be found here.